W3C home > Mailing lists > Public > public-sw-meaning@w3.org > April 2004

Re: How does RDF/OWL formalism relate to meanings?

From: Pat Hayes <phayes@ihmc.us>
Date: Tue, 13 Apr 2004 15:15:24 -0500
Message-Id: <p06001f5bbca1d4d4a7ea@[10.0.100.76]>
To: Graham Klyne <GK@ninebynine.org>
Cc: public-sw-meaning@w3c.org

>Pat,
>
>Being a "a Bear of Very Little Brain",

:-) So, where do you keep your honey?

>I often find myself trying to reduce complex ideas to simple terms. 
>Reading with interest your recent comments on naming and reference, 
>it seems to me that:
>
>(1) "meaning" is something (arbitrary thing?) that is not understood 
>by automata, but some representation of meaning may be transferred 
>by automata if subjected to only meaning-preserving transformations.

I don't know what meaning is. I prefer not to use the word at all. 
Representations, now, I understand what they are. Along with many 
other cogsci folk, my theory of how cognitive agents, including 
people and animals, understand anything is that the understanders 
have representations in their heads which are active in some useful 
way. We can do something similar with software (ie put 
representations in it and manipulate them) so maybe our software can 
be said to understand meanings as well, but the extent to which they 
do so is very, very limited compared to what humans can do with their 
mental representations. (And this idea of mental representations is 
highly debated by some other folk, so best avoided in any SWeb 
standards kind of a forum.)

>(2) formal semantics provides us with tools to describe some 
>representation transformations under which the meaning is preserved 
>(according to the representation used).

Well, OK, though again this talk of preserving meaning as though it 
was a kind of pickle bothers me. What the RDF/OWL operations do is 
preserve truth (or, if you prefer, preserve interpretations). Again, 
truth I understand, but meanings I don't.

>(3) "naming" is a social process,

Yup

>whereby labels are associated with (or "identify") things [a].  All 
>known systems of naming are imprecise.

I wouldnt go so far as to say all. But probably, yes.

>(4) "denotation" (and "interpretation") are formal semantic concepts 
>that are used in the description of meaning-preserving 
>transformations for a representation, which have no formal link with 
>any system of naming.

Well, there  might be some links. For example, it would be perverse 
to say that names don't denote what they name. But there is no 
connection between the actual process of naming, what I called 
baptism, and the formal MT, right.

>  That is, naming is completely opaque as far as formal semantics are 
>concerned.

Right.

>   This is why formal semantics work:  by showing some transformation 
>preserves meaning for any possible system of naming, then it must 
>preserve meaning for any given system of naming.

Right, well put.

>(5) On the web, when the labels used are (a subset of URIs known as) 
>URLs [b] there is a convention for naming based partly on the 
>technical properties of retrieval of representations of web 
>resources.  (While some attempt is being made to improve the 
>situation, the exact operation of this convention does not have a 
>complete rigorous definition, and maybe never will.)

Right. And this relationship, referred to by the word "identify", 
seems to be what the RFC 2396 and TAG architecture text is mostly 
talking about (at least, it makes sense if you interpret it that way.)

BTW, I am trying to figure out how one might actually do a URI baptism. See

http://www.ihmc.us/users/phayes/PatHayes

for a (rather lame) start.


>(6) Currently, there are no widely accepted technical naming 
>mechanisms associated with labels that are not URLs.

Thats the way it seems. Mind you, I still don't fully understand how 
URNs are supposed to work.

>  The only mechanisms are social, in that they rely on out-of-band 
>agreements between people.

Right. Some of the agreements are handled by language, however, so 
there is at least the possibility of getting them handled by 
computational agents if we invent and agree on suitable protocols. We 
can ride on society's coat-tails.

>#g
>--
>
>[a] I use the term "thing" very broadly, in the sense of anything 
>that can be contemplated by a consensual agent (person, etc.).  (The 
>term "consensual agent" comes from a private discussion with an 
>ethical philospher about the nature of trust;  I think its meaning 
>is self-evident, but I can't find any references to back up the 
>sense in which I understand it.)

Well, what would be an example of a non-consensual agent?

>[b] in my view, whether or not a given URI is also a URL is not 
>fixed for all time.   Todays's non-URL may become tomorrow's URL 
>through introduction and consensus concerning new retrieval 
>mechanisms;  e.g. URNs and DDDS.

Ah, good point I had not considered. Hmmm, need to think about that.

Pat

-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes
Received on Tuesday, 13 April 2004 16:15:27 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:42:16 GMT